Strict expected log-product gap for a bounded nontrivial mean-one law
ProvedZeroFreeRRG.Hardcore.bounded_size_bias_strict_gapLet , , and let be a probability law on with mean one, distinct from the point mass at one. There is a coupling of with its size-biased law such that independent copies have a strictly positive expected log-product supermodular gap. The formal statement displays the full gap.
Preamble
import Mathlib set_option autoImplicit false open MeasureTheory Set open scoped Classical BigOperators
Formal statement
theorem ZeroFreeRRG.Hardcore.bounded_size_bias_strict_gap (d : ℕ) (hd : 2 ≤ d) (θ C : ℝ) (hθ : 0 < θ) (hC : 0 ≤ C)
(μ : Measure ℝ) [IsProbabilityMeasure μ]
(hbound : ∀ᵐ x ∂μ, 0 ≤ x ∧ x ≤ C)
(hint : Integrable (fun x : ℝ => x) μ)
(hmean : (∫ x : ℝ, x ∂μ) = 1) (hne : μ ≠ Measure.dirac 1) :
∃ π : Measure (ℝ × ℝ), IsProbabilityMeasure π ∧
Measure.map Prod.fst π = μ ∧
Measure.map Prod.snd π = μ.withDensity (fun x => ENNReal.ofReal x) ∧
let g := fun (xs : Fin d → ℝ × ℝ) (S : Finset (Fin d)) =>
Real.log (1+θ*∏ a, if a ∈ S then (xs a).2 else (xs a).1)
0 < ∫ xs, g xs Finset.univ - g xs ∅ - ∑ i, (g xs {i}-g xs ∅)
∂(Measure.pi fun _ : Fin d => π) := by sorrySource
Chen–Wang–Zhang, Zero-Freeness for Random Regular Graphs up to the Replica Symmetry Breaking Threshold, https://github.com/sigact-stoc/stoc2027/releases/download/papers-1/1461.pdf, proof of Lemma 4.17, pp. 29–31.